Not a silly question at all, actually.
I suspect you could trade level between D10s and Atom to a considerable degree, as both are fairly to very low-noise devices, while your speakers' electronics may not be nearly as much.
For choosing speaker input level, I would first determine the point of the speaker volume control at which output noise begins to audibly rise (no input signal - Atom volume at minimum). Then see whether you can achieve the highest output level you would ever want without any distortion. This may not be feasible but should at least help you dial in levels at the speaker end.
I would set DAC volume to about 80% or so, as I think ESS DACs will hard-clip at 0 dBFS with no headroom. You may not have to do this if you have playback level normalization (e.g. ReplayGain) turned on, as headroom for intersample-overs is generally needed most for material that's quite loud anyway.
Quite honestly, you may well be able to simplify the chain by using a Y splitter cable out of the DAC to connect both Atom and speakers in parallel, thus taking the Atom out of the picture during speaker operation - quite likely to no apparent sonic detriment, while making operation a lot easier (not to mention giving you the ability to choose headphone amp gain and volume freely, only depending on the headphone's needs). Juggling three volume controls in series is in no way user-friendly, especially when you can trade off between the first two to a substantial degree without any apparent effect on the output whatsoever.
Multi-stage volume controls only get you anything when your last amplifier stage is low in noise and easily swamped by noise from the previous stage with volume wide open. A Hi-Fi power amplifier will often qualify, a PC speaker system very well may not. Needlessly adding complexity and inconvenience doesn't make any sense either.